Hi, I'm Stefan!

I'm a Ph.D. candidate in Computer Science at the University of Bonn supervised by Prof. Matthias B. Hullin. Within the scope of my thesis I am working on Differentiable Scene Representation for 4D/6D Capture, Reconstruction and Synthesis.
My interests range from Computer Vision in general to more specific fields like scene representation and reconstruction that can be applied in various use cases such as for games, entertainment, training firefighters over to helping and assessing situations after natural catastrophes.
Feel free to reach out and contact me in case you are interested in my work or want to collaborate on a project.

CV

Publications

* indicates equal contribution


2025

Metamizer: a Versatile Neural Optimizer for Fast and Accurate Physics Simulations

Nils Wandel, Stefan Schulz and Reinhard Klein

International Conference on Learning Representations (ICLR), 2025


2024

Deep Interpolation of Remote Sensing Land Surface Temperature Data with Partial Convolutions

Florian Huber*, Stefan Schulz* and Volker Steinhage

MDPI Sensors, 2024


2023

RoboCup 2022 AdultSize Winner NimbRo: Upgraded Perception, Capture Steps Gait and Phase-Based In-Walk Kicks

Dmytro Pavlichenko, Grzegorz Ficht, Arash Amini, Mojtaba Hosseini, Raphael Memmesheimer, Angel Villar-Corrales, Stefan M. Schulz, Marcell Missura, Maren Bennewitz and Sven Behnke

RoboCup 2022: Robot World Cup XXV, Springer-Verlag, 2023

Projects

Here are some of the projects I am working on and have worked on:

D4Torch

D4torch is a package for equivariant physics based neural networks that work on staggered grids. Standard packages like escnn only work on pixel-grids making them unsuitable for fluid simulations. Thus together with a colleague I am developing a pypi package handling these cases.

Portfolio Website

My personal homepage is now up and running and will be kept up-to-date. I am happy with how it turned out and that it is not some github template. It is used for displaying my work and putting myself out there. Feel free to comment on the design or the contents if you like.